The focus of this paper is on the characterization
of the uncertainties in the evolving states of a diabetic model,
to permit a study of the impact of the time interval between
insulin bolusing and meal initiation on hypo- and hyperglycemic
events. A polynomial chaos based approach is used
to characterize the independent uncertainties in the initial
condition and meal size. Galerkin projection of the resulting
equations reduce the stochastic differential equations to a set of
deterministic equations. This forms the framework to optimize
for the post bolusing time to initiate the meal. Two cost functions
are considered which correspond to the postprandial hypoand
hyperglycemic excursions of the blood glucose. Numerical
results from the minimal Bergman model suggest a 13 and 14
minute interval between bolusing and the initiation of the meal.
more »
« less
Differential Flatness based Run-to-Run Control of Blood Glucose for People with Type 1 Diabetes
The objective of this paper is to develop an open
loop insulin input profile over a span of 24 hours which makes
the glucose trajectory of a Type 1 diabetic person track a target
glucose trajectory. The Bergman minimal model is chosen to
represent the glucose-insulin dynamics which is shown to be
differentially flat. An optimal control problem is posed by
parameterizing the differentially flat output of the Bergman
model using Fourier series, to result in an input profile that
can be repeatedly administered every day. The solution to
the optimization problem is then shown to present acceptable
performance in terms of tracking and adhering to imposed
constraints.
more »
« less
- Award ID(s):
- 1537210
- NSF-PAR ID:
- 10113118
- Date Published:
- Journal Name:
- 2019 American Control Conference (ACC)
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Abstract Background Blood glucose (BG) management is crucial for type-1 diabetes patients resulting in the necessity of reliable artificial pancreas or insulin infusion systems. In recent years, deep learning techniques have been utilized for a more accurate BG level prediction system. However, continuous glucose monitoring (CGM) readings are susceptible to sensor errors. As a result, inaccurate CGM readings would affect BG prediction and make it unreliable, even if the most optimal machine learning model is used. Methods In this work, we propose a novel approach to predicting blood glucose level with a stacked Long short-term memory (LSTM) based deep recurrent neural network (RNN) model considering sensor fault. We use the Kalman smoothing technique for the correction of the inaccurate CGM readings due to sensor error. Results For the OhioT1DM (2018) dataset, containing eight weeks’ data from six different patients, we achieve an average RMSE of 6.45 and 17.24 mg/dl for 30 min and 60 min of prediction horizon (PH), respectively. Conclusions To the best of our knowledge, this is the leading average prediction accuracy for the ohioT1DM dataset. Different physiological information, e.g., Kalman smoothed CGM data, carbohydrates from the meal, bolus insulin, and cumulative step counts in a fixed time interval, are crafted to represent meaningful features used as input to the model. The goal of our approach is to lower the difference between the predicted CGM values and the fingerstick blood glucose readings—the ground truth. Our results indicate that the proposed approach is feasible for more reliable BG forecasting that might improve the performance of the artificial pancreas and insulin infusion system for T1D diabetes management.more » « less
-
The focus of this paper is on the development of an open loop controller for type 1 diabetic patients which is robust to meal and initial condition uncertainties in the presence of hypoand hyperglycemic constraints. Bernstein polynomials are used to parametrize the evolving uncertain blood-glucose. The unique bounding properties of these polynomials are then used to enforce the desired glycemic constraints. A convex optimization problem is posed in the perturbation space of the model and is solved repeatedly to sequentially converge on a sub-optimal solution. The proposed approach is demonstrated on the classic Bergman model for Type 1 diabetic patients.more » « less
-
more » « less
Key points Previous studies in fetuses with intrauterine growth restriction (IUGR) have shown that adrenergic dysregulation was associated with low insulin concentrations and greater insulin sensitivity.
Although whole‐body glucose clearance is normal, 1‐month‐old lambs with IUGR at birth have higher rates of hindlimb glucose uptake, which may compensate for myocyte deficiencies in glucose oxidation.
Impaired glucose‐stimulated insulin secretion in IUGR lambs is due to lower intra‐islet insulin availability and not from glucose sensing.
We investigated adrenergic receptor (ADR) β2 desensitization by administering oral ADRβ modifiers for the first month after birth to activate ADRβ2 and antagonize ADRβ1/3. In IUGR lambs ADRβ2 activation increased whole‐body glucose utilization rates and insulin sensitivity but had no effect on isolated islet or myocyte deficiencies.
IUGR establishes risk for developing diabetes. In IUGR lambs we identified disparities in key aspects of glucose‐stimulated insulin secretion and insulin‐stimulated glucose oxidation, providing new insights into potential mechanisms for this risk.
Abstract Placental insufficiency causes intrauterine growth restriction (IUGR) and disturbances in glucose homeostasis with associated β adrenergic receptor (ADRβ) desensitization. Our objectives were to measure insulin‐sensitive glucose metabolism in neonatal lambs with IUGR and to determine whether daily treatment with ADRβ2 agonist and ADRβ1/β3 antagonists for 1 month normalizes their glucose metabolism. Growth, glucose‐stimulated insulin secretion (GSIS) and glucose utilization rates (GURs) were measured in control lambs, IUGR lambs and IUGR lambs treated with adrenergic receptor modifiers: clenbuterol atenolol and SR59230A (IUGR‐AR). In IUGR lambs, islet insulin content and GSIS were less than in controls; however, insulin sensitivity and whole‐body GUR were not different from controls. Of importance, ADRβ2 stimulation with β1/β3 inhibition increases both insulin sensitivity and whole‐body glucose utilization in IUGR lambs. In IUGR and IUGR‐AR lambs, hindlimb GURs were greater but fractional glucose oxidation rates and
ex vivo skeletal muscle glucose oxidation rates were lower than controls. Glucose transporter 4 (GLUT4) was lower in IUGR and IUGR‐AR skeletal muscle than in controls but GLUT1 was greater in IUGR‐AR. ADRβ2, insulin receptor, glycogen content and citrate synthase activity were similar among groups. In IUGR and IUGR‐AR lambs heart rates were greater, which was independent of cardiac ADRβ1 activation. We conclude that targeted ADRβ2 stimulation improved whole‐body insulin sensitivity but minimally affected defects in GSIS and skeletal muscle glucose oxidation. We show that risk factors for developing diabetes are independent of postnatal catch‐up growth in IUGR lambs as early as 1 month of age and are inherent to the islets and myocytes. -
null (Ed.)Abstract. Engineering wake models provide the invaluable advantage to predict wind turbine wakes, power capture, and, in turn, annual energy production for an entire wind farm with very low computational costs compared to higher-fidelity numerical tools. However, wake and power predictions obtained with engineering wake models can be insufficiently accurate for wind farm optimization problems due to the ad hoc tuning of the model parameters, which are typically strongly dependent on the characteristics of the site and power plant under investigation. In this paper, lidar measurements collected for individual turbine wakes evolving over a flat terrain are leveraged to perform optimal tuning of the parameters of four widely used engineering wake models. The average wake velocity fields, used as a reference for the optimization problem, are obtained through a cluster analysis of lidar measurements performed under a broad range of turbine operative conditions, namely rotor thrust coefficients, and incoming wind characteristics, namely turbulence intensity at hub height. The sensitivity analysis of the optimally tuned model parameters and the respective physical interpretation are presented. The performance of the optimally tuned engineering wake models is discussed, while the results suggest that the optimally tuned Bastankhah and Ainslie wake models provide very good predictions of wind turbine wakes. Specifically, the Bastankhah wake model should be tuned only for the far-wake region, namely where the wake velocity field can be well approximated with a Gaussian profile in the radial direction. In contrast, the Ainslie model provides the advantage of using as input an arbitrary near-wake velocity profile, which can be obtained through other wake models, higher-fidelity tools, or experimental data. The good prediction capabilities of the Ainslie model indicate that the mixing-length model is a simple yet efficient turbulence closure to capture effects of incoming wind and wake-generated turbulence on the wake downstream evolution and predictions of turbine power yield.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
